Synchronous Machine Parameters from Frequency-Response Finite-Element Simulations and Genetic Algorithms

Abstract

This paper presents a novel way to obtain parameters of synchronous machine equivalent circuits from standstill frequency response data using a hybrid genetic algorithm. The genetic algorithm is capable of finding a global minimum within a search interval of the fitness function used to match the equivalent circuit and the measured machine transfer functions, notwithstanding the initial guess of the identification process. Therefore, methods such as the maximum likelihood estimation technique could be substantially enhanced. Results obtained in the identification procedure show that good matching can be obtained with either negative or positive leakage inductance values. These results cast some light on the possible physical meaning that circuit parameters may have. Finite-element modeling is used here to determine the transfer functions of a turbine generator. This approach is consistent with the general aim of obviating the requirement of field testing.